How to write an audio guide script

An audio guide is only as good as its scripts. The technology can be flawless, but if the tracks are long, dense or read like a catalogue entry, visitors stop listening after a few seconds.

Writing for an audio guide is a different craft from writing labels, wall panels or essays. This guide gathers the rules shared by audio guide writers, providers and museum associations in Europe and the United States, with a ready-to-use template, a before-and-after example and a final checklist.

If you are still planning the project as a whole (route, voice, technology, costs), start with How to create an audio guide for a museum.

The golden rule: write for the ear, not the eye

You can reread a label; you cannot re-hear a sentence. People using an audio guide are standing, often moving, surrounded by noise and other visitors. If they lose the thread, they don't go back: they stop listening.

That is why a text meant to be heard works differently from a text meant to be read:

• Less information, not more. For the same length, a good spoken text holds less data than a written one. "Improving" a track rarely means adding facts and dates.
• One step at a time. Ideas should come in sequence, one per sentence, in the order the listener needs them.
• Repetition helps. Coming back to a key idea, or summing up in one sentence, helps anyone whose attention drifted. In an essay it would be a flaw; in audio it is a help.
• Every track must stand on its own. Visitors may play the stops in a different order, skip some, or join halfway through the route.
• The script goes with the eye. An audio guide is like a subtitle for what the visitor has in front of them: it should bring them back to the work, not distract them from it.

The simplest test is to read the script aloud. If you stumble, the narrator will stumble too, and the visitor will be lost.

Before you write: four questions

Professional audio guide writers always start here: what you want to tell, why, to whom and how.

1. Who are you talking to? Enthusiastic adults, families, foreign tourists, school groups. Tone and depth change a lot. If the audience is mixed, write for a curious adult who is not an expert.
2. What is the common thread? An audio guide is not a catalogue. You need a theme that ties the stops together: the history of the place, a character, a question. In a temporary exhibition, it is usually the exhibition's argument.
3. Which message should stay with them? For each stop, write in one line the single idea the visitor should take home. If you can't, the stop has too many topics.
4. What is already in the room? Read the panels and labels so you don't repeat them. Audio should add what the written text cannot: stories, voices, details to look at.

Decide how you will address the visitor, too. In English "you" is the natural choice, but the register can be warm and conversational or more formal. Today's audio guides lean towards the conversational; what matters most is keeping the same register in every track.

Finally, gather your material: catalogue entries, studies, archives, interviews. Look above all for stories: anecdotes, curiosities, people. They are what visitors remember and retell.

How long each track should be

Most tracks should last between 60 and 90 seconds. The sources are almost unanimous: past two minutes, a good share of visitors stop listening.

To turn seconds into words, assume that calm narration runs at roughly 130-150 words a minute (a little slower in German, a little faster in Romance languages). Here are the benchmarks by type of track:

• Stop on a single work: 60-90 seconds, about 130-200 words. One work, one idea, one detail to remember.
• Major work: 90-120 seconds, about 200-260 words. The masterpieces visitors stop in front of anyway.
• Room or section introduction: up to two and a half minutes, up to 330 words. The context before a group of works.
• Secondary stop: 30-45 seconds, about 70-100 words. Minor works, curiosities.
• Transition between rooms: 20-45 seconds, about 40-100 words. Orientation and links between themes.

Three practical rules complete these benchmarks:

• Vary the length. A run of identical tracks becomes monotonous. Alternate short and longer ones.
• Split rather than stretch. If a work deserves more, offer one or two optional deep dives: those who want them listen, the others move on.
• Add up the total. The overall length should stay around half of the average visit time. A guide with 20 stops of 90 seconds is already half an hour of audio, before counting the time spent walking and looking.

If you plan translations, aim the source text at the lower end: many languages "grow" in translation, and the same track could run 20% longer.

The five-part structure of a track

A fixed structure helps writers work fast and listeners find their way. The one that recurs in almost every source has five parts:

1. Hook (1 sentence, 5-10 seconds). A question, a surprising detail, an image. This is when visitors decide whether to keep listening: don't waste it repeating the title, artist and date they have just read on the label.
2. Look (1-2 sentences). Direct the eye to what can be seen: a gesture, a material, a colour, a size. Even people with good eyesight need to know where to look.
3. Meaning (2-4 sentences). The story, the context, why this work matters. One topic only: don't pile up biography, technique, period and provenance.
4. The detail to remember (1-2 sentences). A concrete image the visitor will carry into the next room, or tell someone about.
5. Close (1 sentence). Signal that the track is over and, if useful, invite one more look or point to the next stop.

The weakest point of audio guide scripts is almost always the opening. A start like "This painting, made in 1532 by..." sounds like a catalogue entry. It is far better to begin with something the visitor can notice straight away.

Template to copy

• Stop number and title:
• Target length: 60-90 seconds
• Key idea (1 line, internal use):
• Hook:
• Look:
• Meaning:
• Detail to remember:
• Close:
• Optional deep dive:

Language and style: the practical rules

Sentences

• Short: roughly 12-18 words, one idea per sentence. If a sentence needs a semicolon, it is too complex for listening.
• Active: "The painter painted..." is easier to follow than "It was painted by...".
• Subject and verb close together: avoid long asides that force the listener to remember how the sentence began.
• Varied: mix short sentences with slightly longer ones, so the rhythm doesn't become mechanical.

Words

• Concrete and everyday: "the clay was left rough, and you can still see the fingerprints" works better than "the work expresses a dialogue with matter".
• Technical terms: if you need them, explain them in a few words the first time; otherwise cut them. Precision does not mean specialist jargon.
• No judgements on the visitor's behalf: instead of "you'll be amazed by its size", give a measure people can picture: "the vase is almost as tall as a ten-year-old child".
• Avoid "as you can see": "look at", "find", "notice" work better.

Numbers, dates and names

• One date per track, two at most. Nobody remembers a string of years.
• Round and compare: "about five hundred years ago" or "in the early sixteenth century" is easier to grasp than "in 1527". Leave exact measurements to the panels.
• Proper names: few per track, always spelled the same way, with pronunciation notes for the narrator.

Tone

Write the way a knowledgeable friend would talk, not like a manual. Writers with twenty years of audio guides behind them recommend showing your own enthusiasm: saying what surprises you about a work makes the track more alive. Always respect the listener's intelligence, though: simple does not mean childish, unless the guide is designed for children.

Guiding the eye and telling stories

Directions anchored to the work, not to the visitor. "Look to your left" only works if the visitor is exactly where you expect. In a crowded room that almost never happens. Better: "on the right side of the painting, the figure holding the lantern". If you need to give directions, use visible landmarks: "with your back to the lift, walk towards the large statue".

Describe first, then explain. A short description of what can be seen ("a small bronze figure, kneeling, arms raised") gives the listener something concrete to hold on to. It helps everyone, and it is essential for blind and partially sighted people.

Stories beat dates. People remember anecdotes, curiosities and characters far better than abstract information. An object of little artistic importance can become unforgettable thanks to its story. Ask yourself: what would an experienced guide tell a friend in front of this work?

Questions make the visitor active. "What do you notice about the light in this room?" followed by a short pause turns listening into looking.

More voices make the guide more alive. Short contributions from the curator, the artist, a conservator or the director add authenticity and take minutes to record. Quotations from letters, diaries or archive documents, read by a second voice, also work well.

Sound and music, in moderation. A well-chosen sound (the clatter of a loom next to a textile, music from the period) makes the content more sensory. Used everywhere, it tires the listener and covers the voice. Always check that speech stays clear even in a noisy room.

Preparing the script for the voice and for translation

The final text is not just prose: it is a working script for the narrator (or the synthetic voice), the translator and whoever edits the audio. Preparing it well avoids improvising in the studio and costly corrections.

What to include in each track's script

• the stop number, identical to the one displayed next to the work and to the audio file name;
• the title of the work and its location in the museum;
• the target length;
• the text to be read;
• pronunciation notes for proper names, places and foreign terms;
• pauses and words to stress, where needed;
• cues for music or sound;
• who approved the text.

If you use a synthetic voice, listen back to every track: AI voices often get exactly the proper names, dates and acronyms wrong. Write numbers the way they should be spoken ("fifteen twenty-seven" or "in the sixteenth century") and, if the tool allows it, add a phonetic spelling for difficult names.

If the text will be translated:

• avoid puns, idioms and cultural references that don't translate; if they are essential, flag them for the translator;
• always use the same name for people, places and concepts, and prepare a short glossary shared by all languages;
• give the translator the target length too, not just the word count;
• machine translation and AI are fine for a first draft, but the main languages need a review by a native speaker, especially on sensitive historical topics.

Accessibility. Provide a written transcript of every track. If your audience includes blind or partially sighted people, the best solution is a parallel version with fuller visual descriptions. For children, people with cognitive difficulties or visitors short on time, you can prepare an "easy" version: the same content, shorter sentences, fewer names and dates.

Example: before and after

The example concerns an imaginary object: a ceramic jug kept in a small civic museum.

Before: a text written to be read

Polychrome maiolica jug, central Italian manufacture, second half of the 16th century, height 34 cm. The work, which entered the museum in 1923 as a bequest from the Rossi family, features stylised plant motifs in cobalt blue and antimony yellow, characteristic of contemporary production, and attests to the spread of Renaissance taste even in provincial domestic settings.

One very long sentence, five technical facts, no invitation to look.

After: a text written to be heard (about 45 seconds)

Look at the handle of this jug. In one spot the glaze has almost gone, worn away by a hand that gripped it thousands of times.

We are in the second half of the sixteenth century. This was not a work for display: it was an everyday object, used to bring water or wine to the table.

And yet whoever painted it took care. Follow the blue leaves that run around the belly of the jug: every one of them is different.

The Renaissance, in other words, was not only in the palaces of the powerful. It found its way into kitchens too.

Before you move on, find the mark on the handle again. It is the oldest trace of the people who used it.

Same object, fewer facts: but the visitor knows where to look and takes an image home.

Checklist before recording

• Every track has a single key idea that fits in one line.
• The length is between 60 and 90 seconds (or justified if longer).
• The first sentence grabs attention and doesn't repeat the label.
• The text says where to look, with references to the work rather than to the visitor.
• Every track makes sense when heard on its own.
• At most one or two dates and a few proper names per track.
• Technical terms are explained or removed.
• The text has been read aloud and timed.
• A colleague outside the subject has heard it and understood it first time.
• An expert has checked the accuracy of the content.
• The register (conversational or formal) is the same in every track.
• Pronunciation notes for the narrator or synthetic voice are ready.

Common mistakes

• Reading out the catalogue entry instead of writing a new text.
• Trying to say everything: too many topics in the same track.
• Starting from the data (artist, date, technique) instead of what can be seen.
• Directions like "on your left" that stop working when the room is full.
• References to earlier tracks ("as we saw before") that confuse anyone who skipped a stop.
• Approving the text without ever hearing it in the room, with the real noise.

Frequently asked questions

How many words should an audio guide track have?

For a 60-90 second stop, roughly 130 to 200 words. Stay towards the lower end if the text will be translated.

What is the difference between a label and an audio guide script?

A label is read and can be reread; audio is heard only once. It needs shorter sentences, fewer numbers, a more direct tone and directions on where to look.

Should the tone be conversational or formal?

Both work. A conversational tone is more direct and engaging, a formal one more reserved. What matters most is consistency across all tracks.

Can AI write audio guide scripts?

It can help prepare a first draft from catalogue entries. But it must always be reviewed by someone who knows the works: AI tends to oversimplify, invent details and get exactly the lesser-known facts wrong.

How do you write for children?

With shorter tracks, questions, small observation challenges ("find the dog hidden in the painting") and language that is simple but not childish. It is worth creating a dedicated route rather than adapting the adult one.

Should I include the curator's voice?

Whenever possible, yes. Even 20-30 seconds of direct contribution give the audio guide an authenticity no script can imitate.